Study On Short-term Power Load Prediction Based On Support Vector Regression Modeling Method

With the speeding up of electric power intelligentize progress, especially it is strengthened in study of power load information change. It is great significance to improve forecasting accuracy of short-term power load and promote the construction and development of smart grid.Currently there are several kind of prediction methods to use in power load short-term forecasting, but it is difficult to make short-term power load forecasting due to the limits of application condition of prediction model, therefore, in this study we chose the support vector regression (SVR) modeling method to forecast. Based on the structural risk minimization principle of statistical theory of support vector regression, the nonlinear problem of original space was transformed into the feature space by using kernel function mapping, and can realize the high efficiency machine learning under small sample.In study, firstly we used the SVR modeling method to predict the classical Henon chaotic time series, and verified indirectly the feasibility of short-term load forecasting. Then, we combined with the historical power load data of economic development zone of some city in Guizhou province to predict the short-term power load by using the SVR modeling method. In the meantime we also selected the artificial neural network(ANN) method to predict the same set of data, it shows that the SVR method has better short-term power load prediction accuracy than the ANN modeling method. It indicate that the SVR is a good way to realize power load short-term load forcasting.